Research On Text Big Data Mining Of Stock Prediction

In the era of big data, the value of data received extensive attention of various fields. Mining the potential value from the huge amount of unstructured text data, becomes the main driving force of Big Data processing research.With the development of mobile Internet technology, the amount of network data increases to an exponential trend. Investors can always exchange investment experience over the network. However, in order to dig for investor sentiment and assist investors in scientific decision-making and effective investment combined with the specific application requirements, from the perspective of the Big Data processing system architecture design, this dissertation mainly studies its related models and key technologies that support this implementation, and finally implements the support Big Data text processing system to forecast the stock price. The mainly work of this dissertation are as follows:1) The design of the system architecture and key technology research. Firstly, study the Big Data acquisition technology, data storage technology and data mining technology. Secondly, compared with the traditional way of text mining, data mining and text classification process are optimized. The algorithm complexity and the amount of data in the process of mining and processing are reduced. Finally, through the Big text Data architecture design and processing support summarized the key technologies to implement this architecture.2) The irrelevant data collection is reduced. This dissertation study mainly through the focused crawler based on ontology.By building stock ontology, computing topic pages degree guidance on topics related to reptile high priority crawling pages, reduced the irrelevant data collection.3) Multiple source stock forecast model is designed. Based on large data processing technology from investor sentiment in the text mining, to calculate the investor sentiment index. With the analysis of factors that affecting the stock price, and extraction of characteristics of the main ingredients. Combined with support vector machine regression problems in processing function aspect superiority, the multiple source stock forecast model is designed, and the validity of the model is verified by experiment.By mining investor sentiment through Big Data processing technology, and analyzing the various factors that affect stock prices, a multiple source stock forecast model is designed. This model is a prototype to implement the stock trend forecast system.